Introduction

• C. difficile infection (CDI) can lead to adverse outcomes including recurrent infections, colectomy, and death (1).
• The composition of the gut microbiome plays an important role in determining colonization resistance and clearance when exposed to C. difficile (2, 3).
• Identifying the specific microbiome features that distinguish severe CDI cases would eventually allow clinicians to tailor interventions based on a patient’s risk, ultimately leading to better health outcomes.

Dataset

• We have 16S amplicon sequence data from 1191 CDI patient stool samples, with cases classified as severe or not severe according to three separate definitions:
  • the Infectious Diseases Society of America (IDSA) definition: white blood cell count ≥ 15 k/μL and serum creatinine level ≥ 1.5 mg/dL (4).
  • the CDC definition: ICU admission, colectomy, or death occurring within 30 days of CDI, and confirmed as attributable to CDI via clinical chart review.
  • all-cause severity: ICU admission, colectomy, or death occurring within 30 days of CDI, regardless of the cause.

Methods

• Sequences were processed with mothur according to the MiSeq SOP and clustered into de novo OTUs at a 3% distance threshold (5, 6).
• We then trained machine learning (ML) models with OTU abundances as features to predict the IDSA severity, CDI-attributable severity, and all-cause severity of CDI cases using the mikropml R package (7, 8).
• The dataset was randomly split into training and testing sets with 80% of the data in the training set, then random forest models were trained with 5-fold cross-validation repeated 100 times, and performance as the area under the receiver-operator curve (AUROC) was measured on the testing set for the best model.
• This was repeated for 100 different random partitions of the data into training and testing sets.

Results

• This process yielded median AUROC values of TODO
• Feature importance was determined with a permutation test for the best random forest model, revealing that the top 5 OTUs that contributed the most to model performance were TODO.

Conclusions

• The modest performance may be improved in future work by training to predict clinically confirmed adverse patient outcomes rather than IDSA severity, such as recurrence, admission to intensive care, colectomy, or death.
• Predicting a patient’s risk of experiencing a severe CDI and identifying the specific microbiome features that distinguish severe CDI cases will allow clinicians to tailor interventions based on each patient’s individual microbiome, ultimately leading to better health outcomes.

Acknowledgements

This research was supported by National Institutes of Health grants U01AI124255 and the Michigan Institute for Clinical and Health Research Postdoctoral Translational Scholars Program (UL1TR002240 from the National Center for Advancing Translational Sciences).